基于视觉特征的农业机械三维模型相似性评价

The Similar Evaluation on the 3D Model of Agricultural Machinery Based on Visual Feature

目前已有多种理论、方法应用到三维模型的相似性评价中,但是通过对三维模型进行投影,将其转化为二维图像,然后通过图像匹配方式间接实现对模型的评价在农机产品CAD模型领域却少有应用。因此,本文以农业机械零件三维模型的投影视图为研究对象,提出了一种两阶段检索方法,首先以傅里叶描述子作为过滤方法对图像边缘特征提取比较,进行初期匹配,之后以极半径不变矩对图像的整体区域进行二次精确评价,分别计算图像描述子的距离,最终实现农业机械模型的相似性评价。结果表明,算法检索性能取得了良好的检索效果,能够满足实际要求。

There are various theories and methods applied to evaluate the similarity of 3D models until now. However,converting them into 2D images and achieving matching of image to realize the assessment are hardly adopted in specific CAD fields. This paper proposed a two-phases retrieval method based on agricultural mechanical parts’ projection views aimed at the situation. Firstly, we adopted Fourier descriptor to extract edge information as initial matching and then calculate features of whole area by the use of polar radius of invariant moment for accurate evaluation after computing the difference between image descriptors, finally, we realized the similarity evaluation of agricultural mechanical parts. Test results showed that our algorithm had good retrieval performance and achieved great efficiency, so it could satisfy the actual requirements.